采用了四种聚类方法对降糖类药物拉曼光谱进行了快速、 无损判别。 采集九种降糖类药品共48个样品的拉曼光谱, 经截波、 基线校正、 平滑、 矢量归一化等预处理后, 分别采用K-均值、 系统聚类法、 自组织图(SOM)及PCA-SOM四种不同聚类方法做聚类判别。 结果表明, 自组织图与K-均值、 系统聚类法相比, 聚类结果较好, 并且SOM结合PCA后的PCA-SOM结果最优。 为降糖类药品的快速判别从聚类的角度提供了一种新的方法。

In the present paper, four kinds of cluster analysis methods were used in rapid, non-destructive discrimination of hypoglycemic tablets by the Raman spectroscopy technology. Nine kinds of hypoglycemic tablets, including 48 samples, were determined using a Raman spectrometer. The sample data were pretreated with the methods of frequency range cutting, baseline correction, smoothing and vector normalization, then were analyzed by K-means, hierachical cluster, self-organizing maps(SOM) and PCA-SOM respectively. The results demonstrated that SOM was better than K-means and hierachical cluster, and it provided the best discrimination when combined with PCA. The research offers a new approach to the rapid discrimination of different kinds of hypoglycemic tablets.